- Sparkle's dense PCIe 5.0 design enables lower-cost access to artificial intelligence and data exploration.
- Up to 32 Intel GPUs provide 768 GB of video memory for scalable AI computing.
- The 10,800W redundant power system and advanced cooling system enable continuous heavy workloads.
Taiwanese graphics card maker Sparkle has unveiled its new dense, scalable GPU server that could offer an alternative to expensive AI accelerator systems such as NvidiaH100, H200 and B200 GPUs for the data center.
The C741-6U-Dual 16P is designed to support up to 16 Arc Pro B60 Dual graphics cards, for a total of 32 Intel GPUs and as much as 768 GB of video memory.
Each Arc Pro B60 Dual is equipped with two Battlemage BMG-G21 GPUs, and when fully loaded the system will have a whopping 81,920 GPU cores.
Improved Cooling
The server supports 24GB and 48GB GPU models when paired with 4th or 5th generation Intel Xeon Scalable processors.
It uses the Intel C741 chipset and supports up to 32 DDR5 memory slots, allowing for 256GB per module for intensive workloads.
Sparkle built the system around a 6U chassis with massive power and cooling capabilities.
The top configuration uses five 2,700W titanium power supplies with a total power output of up to 10,800W, while the lower version uses four 2,400W power supplies with a power rating of 7,200W.
The system, which features up to 15 fans, is clearly designed to efficiently manage heat during continuous workloads.
Each GPU connects to the processor via a PCIe 5.0 x8 interface directly or through a Microchip switch, maximizing data throughput and avoiding potential bottlenecks caused by shared links.
There is support for two M.2 slots and several NVMe and SATA bays for local storage.
While it's impossible to know at this stage how the Arc Pro B60 will perform in large-scale inference or training workloads, the flexibility offered by the hardware may appeal to researchers and developers who need affordable parallel computing power.
Unsurprisingly, Sparkle has not revealed the price of the C741-6U-Dual 16P, but potential buyers can purchase request on the website.
By using Video card
Follow TechRadar on Google News. And add us as your preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the “Subscribe” button!
And of course you can also Follow TechRadar on TikTok for news, reviews, unboxing videos and get regular updates from us on whatsapp too much.